Using NLP on unstructured EHR for COVID-19 surveillance
We applied NLP pipeline and developed association analysis called ConceptWAS to track early symptoms for COVID-19
We developed machine learning and deep learning models to predict 10-year CVD risk using longitudinal EHR and genetic data.
Deep phenotyping using unsupervised machine learning
Applied a state-of-arts tensor factorization method to longitudinal EHR data